When we started building our project, we instinctively went all in on the cloud: Google Kubernetes Engine, managed services, an architecture ready for future scale. On paper, everything made sense—Kubernetes, high availability, automation. The problem was that at the time our app was just an MVP: a small number of users, a small team, and fast iterations.
We were hosting the frontend, backend, databases, and test environments. Nothing more. Yet we were paying for infrastructure as if we were handling traffic at the level of a mature product. Over time, it became clear that the cost and complexity were completely misaligned with the real needs of the project.
Instead of “optimising the cloud”, we took a step back. We migrated the infrastructure to a VPS. Kubernetes stayed. Order in deployments stayed too, thanks to GitOps and Argo CD. What disappeared was the entire layer of managed services that we didn’t actually need. On a single, understandable setup, we ran the frontend, backend, databases, and test environments.
The result was unambiguous: infrastructure costs dropped by roughly thirty times.
Beyond the business and cost arguments, there’s also something simply enjoyable about it. Setting up your own VPS, running Kubernetes on it, and hosting your frontend, backend, or databases gives you a real sense of control over the system. From time to time, it’s worth going back to these fundamentals, if only to remember what’s really happening under the hood. :)